TSMC Has Bought ASML’s High-NA EUV Machines, But Here’s Why They’re Staying in the Lab

This position clarifies confusing headlines from April 2026, when TSMC Deputy Co-COO Kevin Zhang said the company had "no current plans" to adopt High-NA EUV and called the machines "very, very expensive" . At the shareholder meeting, Wei walked back the impression that TSMC was permanently opting out. The real strategy, he explained, is to extend the performance of existing low-NA EUV tools for as long as possible. TSMC's R&D team continues to find ways to scale and improve chip performance without switching to the next-generation equipment, a capability that Zhang had separately called "amazing" .

Industry analysts now expect TSMC’s A13 node, targeted for 2029, will not require High-NA EUV, pushing the mass-production debut for the technology at TSMC to some point beyond that timeframe .

The Competitive Landscape: Who’s Moving First

Wei's deliberate pace stands in contrast to the aggressive moves by Intel and SK Hynix. Both companies are preparing to use High-NA EUV tools for advanced AI logic and high-bandwidth memory chips as early as 2027 . Intel, which received the industry's first High-NA EUV system in December 2023, has already exposed roughly 300,000 wafers on its development machines as it prepares for its 14A node . SK Hynix became the first chipmaker to assemble a High-NA tool for R&D at a memory facility .

ASML CEO Christophe Fouquet added pressure to the timeline in May 2026, stating that the first commercial chips produced on High-NA EUV machines—spanning both memory and logic products—would arrive “within months” .

Wei dismissed any suggestion that early movers are gaining a permanent advantage. He said TSMC has “never lacked competitors” and will continue to rely on process innovation and existing EUV optimization rather than rushing into an expensive tool transition . He specifically cited Intel’s 18A process and Samsung Foundry as competitive pressures, but expressed confidence that TSMC will “keep working to stay ahead” .

AI Demand Is Outpacing Supply, and It Won’t Let Up

Beyond the equipment strategy, Wei delivered a stark message about the artificial intelligence boom: “AI demand is so high, we can only produce so much.” He told shareholders that demand for advanced semiconductors driven by AI will remain supply-constrained for years to come .

Demand is coming from consumers, enterprises, and sovereign AI programs across governments, and Wei said customers remain overwhelmingly optimistic about the AI industry’s trajectory . Even with new manufacturing capacity, including ongoing factory expansions in the United States, TSMC cannot fulfill all the orders it is receiving .

Financial Results Reflect the Boom

The AI surge is flowing directly into TSMC’s financials. The company reported Q1 2026 revenue of NT$1.13 trillion ($35.7 billion), a 35% increase year-over-year, and net profit jumped 58% to a record T$572.5 billion ($17.7 billion) . Wei guided for full-year 2026 revenue growth of more than 30% .

These figures follow a record 2025, when revenue and earnings per share both hit all-time highs and the stock price more than doubled, climbing from NT$950 to NT$2,425 per share in a single year .

To keep pace with demand, TSMC plans 2026 capital expenditures of $52 billion to $56 billion, much of it directed at expanding advanced-node manufacturing capacity .